This MCP server demonstrates a modular, context-aware client-server architecture built with Spring Boot and Spring AI, integrated with Azure OpenAI models. The project consists of two Maven modules: a chat client and an AI-powered server. The client exposes a /chat endpoint and forwards prompts to the server, propagating headers and context. The server registers and executes AI-powered tools (such as user management) and propagates context across asynchronous operations, enabling distributed tracing and consistent user experiences. To use it, first ensure you have Java 17+, Maven, and an Azure OpenAI key and endpoint. After building, run the server and client to see a collaborative chat workflow where the client queries AI-powered tools via the server, with context preserved across threads and requests.

Prerequisites:

• Java 17+
• Maven 3.8+
• Azure OpenAI API key and endpoint

Install steps:

1. Clone the repository git clone <repo-url> cd mcp-client-server-azure

2. Build the multi-module project mvn clean install

3. Run the server (from the project root or using the Maven wrapper) cd server ./mvnw spring-boot:run

   server runs on http://localhost:8081 by default

4. Run the client (in a new terminal) cd ../client ./mvnw spring-boot:run

5. Open the client API and test the chat flow at http://localhost:8080/chat (or the configured ports for client).

Configuration notes:

• Update client properties with your Azure OpenAI credentials in client/src/main/resources/application.properties
• Update server properties as needed in server/src/main/resources/application.properties

Tips and caveats:

• Ensure your Azure OpenAI API key and endpoint are correctly configured in the client properties (spring.ai.azure.openai.api-key and spring.ai.azure.openai.endpoint).
• The system propagates context (headers, MDC, locale, and request attributes) across async boundaries, which is crucial for observability and debugging.
• If you encounter port conflicts, adjust server.port in server/src/main/resources/application.properties.
• The architecture supports dynamic tool registration and context-aware responses; explore the server's UserService endpoints for CRUD operations and search.
• For containerized usage, you can adapt the Docker image name to your own registry and adjust volumes/secrets for Azure credentials.
• If you modify context propagation components, validate that headers and MDC values propagate through async task execution and tool calls.